WO2024109528A1 - 虚拟对象的控制方法、装置和存储介质及电子设备 - Google Patents

虚拟对象的控制方法、装置和存储介质及电子设备 Download PDF

Info

Publication number
WO2024109528A1
WO2024109528A1 PCT/CN2023/129848 CN2023129848W WO2024109528A1 WO 2024109528 A1 WO2024109528 A1 WO 2024109528A1 CN 2023129848 W CN2023129848 W CN 2023129848W WO 2024109528 A1 WO2024109528 A1 WO 2024109528A1
Authority
WO
WIPO (PCT)
Prior art keywords
virtual object
operation information
operation mode
information
user
Prior art date
Application number
PCT/CN2023/129848
Other languages
English (en)
French (fr)
Inventor
刘行
刘章术
Original Assignee
腾讯科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 腾讯科技（深圳）有限公司 filed Critical 腾讯科技（深圳）有限公司
Publication of WO2024109528A1 publication Critical patent/WO2024109528A1/zh

Links

Classifications

    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
    • A63F13/00—Video games, i.e. games using an electronically generated display having two or more dimensions
    • A63F13/55—Controlling game characters or game objects based on the game progress
    • A63F13/56—Computing the motion of game characters with respect to other game characters, game objects or elements of the game scene, e.g. for simulating the behaviour of a group of virtual soldiers or for path finding

Definitions

  • the present application relates to the field of computers, and in particular to a control method, device, storage medium and electronic device for a virtual object.
  • virtual objects controlled by users interact with virtual objects controlled by AI to gain experience, materials, or rewards for completing levels.
  • related technologies usually adjust the strength of AI by simply adjusting the values of virtual objects controlled by AI to avoid the problem of a large gap between the strength of AI and the level of the players.
  • the embodiments of the present application provide a method, device, storage medium and electronic device for controlling a virtual object, so as to at least solve the technical problem of low efficiency in controlling a virtual object.
  • a method for controlling a virtual object comprising:
  • a first virtual object and a second virtual object participating in the cloud game are displayed, wherein the first virtual object is a virtual object controlled by a user of the cloud game, and the second virtual object is a virtual object controlled by artificial intelligence;
  • Acquire first operation information generated by the user on the first virtual object during the running of the cloud game determine a first operation mode corresponding to the second virtual object based on the first operation information, and control the second virtual object according to the first operation mode;
  • the first operation mode corresponding to the second virtual object is adjusted to a second operation mode based on the second operation information, and the second virtual object is controlled according to the second operation mode, wherein the first operation information is different from the second operation information, and the first operation mode is different from the second operation mode.
  • a control device for a virtual object including:
  • a first display unit is used to display a first virtual object and a second virtual object participating in a cloud game during the running of the cloud game, wherein the first virtual object is a virtual object controlled by a user of the cloud game, and the second virtual object is a virtual object controlled by artificial intelligence;
  • a first determining unit configured to obtain first operation information generated by the user on the first virtual object during the running of the cloud game, determine a first operation mode corresponding to the second virtual object based on the first operation information, and control the second virtual object according to the first operation mode;
  • the first adjustment unit is used to adjust the first operation mode corresponding to the second virtual object to a second operation mode based on the second operation information when second operation information generated by the user on the first virtual object during the running of the cloud game is obtained, and control the second virtual object according to the second operation mode, wherein the first operation information is different from the second operation information, and the first operation mode is different from the second operation mode.
  • a computer program product or a computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium.
  • a processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, so that the computer device executes the control method of the virtual object as described above.
  • an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the virtual object control method through the computer program.
  • cloud gaming is used to realize real-time collection of operation information, and then the operation mode corresponding to the virtual object controlled by artificial intelligence is determined based on the collected operation information. Then, based on the real-time changes of the above operation information during the cloud gaming process, the operation mode corresponding to the virtual object controlled by artificial intelligence is flexibly adjusted, thereby achieving the purpose of updating the operation mode of artificial intelligence according to the real-time game level of the user, thereby achieving the technical effect of improving the control flexibility of virtual objects, and thus solving the technical problem of low control flexibility of virtual objects.
  • FIG1 is a schematic diagram of an application environment of a method for controlling a virtual object according to an embodiment of the present application
  • FIG2 is a schematic diagram of a process of a method for controlling a virtual object according to an embodiment of the present application
  • FIG3 is a schematic diagram of a method for controlling a virtual object according to an embodiment of the present application.
  • FIG4 is a schematic diagram of another method for controlling a virtual object according to an embodiment of the present application.
  • FIG5 is a schematic diagram of another method for controlling a virtual object according to an embodiment of the present application.
  • FIG6 is a schematic diagram of another method for controlling a virtual object according to an embodiment of the present application.
  • FIG7 is a schematic diagram of another method for controlling a virtual object according to an embodiment of the present application.
  • FIG8 is a schematic diagram of another method for controlling a virtual object according to an embodiment of the present application.
  • FIG9 is a schematic diagram of another method for controlling a virtual object according to an embodiment of the present application.
  • FIG10 is a schematic diagram of another method for controlling a virtual object according to an embodiment of the present application.
  • FIG11 is a schematic diagram of a control device for a virtual object according to an embodiment of the present application.
  • FIG. 12 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application.
  • Artificial Intelligence is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
  • artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence.
  • Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making.
  • Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies.
  • Basic artificial intelligence technologies generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operating/interactive systems, mechatronics, and other technologies.
  • Artificial intelligence software technologies mainly include computer vision technology, speech processing technology, natural language processing technology, and machine learning/deep learning.
  • artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.
  • Cloud gaming also known as gaming on demand, is an online gaming technology based on cloud computing technology. Cloud gaming technology enables thin clients with relatively limited graphics processing and data computing capabilities to run high-quality games.
  • the game is not played on the player's game terminal, but on a cloud server, which renders the game scene into a video and audio stream and transmits it to the player's game terminal over the network.
  • the player's game terminal does not need to have powerful graphics computing and data processing capabilities, but only needs to have basic streaming media playback capabilities and the ability to obtain player input commands and send them to the cloud server.
  • a method for controlling a virtual object may be, but is not limited to, applied in an environment as shown in FIG. 1 .
  • the method may include, but is not limited to, a user device 102 and a server 112 .
  • the user device 102 may include, but is not limited to, a display 104 , a processor 106 , and a memory 108 .
  • the server 112 includes a database 114 and a processing engine 116 .
  • the specific control process may include the following steps:
  • Step S102 the user device 102 obtains first operation information of a user on the first virtual object from a client corresponding to the first virtual object 1002;
  • Steps S104-S106 sending the user's first operation information on the first virtual object to the server 112 via the network 110;
  • Step S108 the server 112 determines the first operation mode corresponding to the second virtual object 1004 based on the first operation information through the processing engine;
  • Steps S110-S112 send the second operation information of the first operation mode corresponding to the second virtual object 1004 to the user device 102 through the network 110, the user device 102 processes the second operation information of the first operation mode through the processor 106, and displays the process of controlling the second virtual object 1004 to release skills based on the second operation information of the first operation mode on the client, and stores the first operation information and the second operation information in the memory 104.
  • Step S114 the user device 102 sends a prompt mark of location information to the device where the second virtual object 1004 is located.
  • the above steps may be completed independently by the client or the server, or by the client and the server working together, such as the user device 102 executing the above steps S108, thereby reducing the processing pressure of the server 112.
  • the user device 102 includes but is not limited to a handheld device (such as a mobile phone), a laptop computer, a desktop computer, a vehicle-mounted device, etc., and the present application does not limit the specific implementation of the user device 102.
  • the method for controlling a virtual object includes steps S202-S206:
  • the first operation mode corresponding to the second virtual object is adjusted to the second operation mode based on the second operation information, and the second virtual object is controlled according to the second operation mode, wherein the first operation information is different from the second operation information, and the first operation mode is different from the second operation mode.
  • the control method of the virtual object can be applied to, but not limited to, cloud gaming scenarios.
  • Cloud gaming can be understood as, but not limited to, game players inputting commands through terminal devices, while the cloud server is directly responsible for real-time rendering of game animation effects, graphics computing, and data processing, which greatly reduces the computing requirements of game players’ terminal devices. Request. For traditional games, this part of the work is usually done by the host of the game player's terminal device, and the large amount of computing power required requires a large and expensive terminal device to execute. In cloud gaming mode, since all graphics calculations and game scene rendering are separated from the local hardware, the game player's terminal device only needs to be responsible for display and encoding functions, and does not require high power consumption and storage space.
  • the AI strength is usually adjusted by simply adjusting the numerical value of the virtual object controlled by AI. If the adjusted AI strength is too high or too low, it will cause disgust among game players, thereby greatly reducing the fun of the game. Therefore, the related art has a technical problem of low flexibility in adjusting AI strength.
  • the first virtual object can be understood as, but not limited to, a virtual object controlled by a user (i.e., the current game player), for example, the user can control the movement, challenge, release skills, etc. of the virtual object, and no unnecessary restrictions are made here.
  • the second virtual object can be understood as, but not limited to, a virtual object controlled by an artificial intelligence simulation user, where the user refers to a human being, that is, the artificial intelligence simulates human thinking (specifically, the thinking of the game player) to control the virtual object.
  • the relationship between the first virtual object and the second virtual object can be, but not limited to, belonging to the same camp or hostile camp of the game, for example, the first virtual object controlled by the user and the second virtual object controlled by the artificial intelligence simulation user are in a hostile relationship of different camps.
  • the game background can comprehensively consider the user's game level based on factors such as the user's experience level, historical winning rate, operation score, etc., and based on the user's game level, the artificial intelligence controls the second virtual object to perform corresponding operations, thereby achieving the purpose of determining the operation mode simulated by the artificial intelligence according to the comprehensive game level of the game player, and achieving the technical effect of improving the accuracy of the operation mode determination.
  • the game mode of the second virtual object can be predetermined based on, but not limited to, the user's historical information and experience level.
  • the first operation information generated by the user for the first virtual object is obtained in real time, such as whether the user completes an operation instruction with a high difficulty (for example, reaching a preset difficulty threshold), the user's operation speed, the number of challenges the user has made, the number of times the user has been defeated, etc.
  • the operation mode corresponding to the second virtual object controlled by artificial intelligence (also called the operation mode of artificial intelligence) is determined based on the first operation information.
  • the level of artificial intelligence is adjusted to a high-difficulty level according to the first operation information.
  • the difficulty level matches the difficulty of the user's operation, thereby achieving the purpose of determining the operation mode of artificial intelligence according to the user's first operation information, and then generating an AI with moderate difficulty, which brings a better game experience to the player.
  • the operation mode may be, but is not limited to, an operation mode determined by the degree of difficulty, such as a novice mode, a normal mode, and a challenge mode in a game, etc., or an operation mode determined by the functional purpose of a game section, such as an entertainment mode, a formal mode, etc. in a game, or different operation modes determined according to the gameplay of the game player. For example, in a game that includes collection, confrontation, and other gameplay, the player is more inclined to the collection gameplay, so the operation mode of the AI can be adjusted to an AI with a higher collection level (for example, reaching a preset level) according to the player's gameplay.
  • the user likes to collect skins of various props, but the confrontation shooting level is poor.
  • the artificial intelligence can obtain the user's gameplay and adjust the operation mode corresponding to the virtual object controlled by it to an operation mode with better dressing effects but poor confrontation level.
  • the relevant technology usually determines the operation mode of AI according to the player's experience value, and does not take into account the player's game play. For example, player A prefers to dress up virtual objects rather than operate virtual objects to attack. Even if player A's overall experience value is high, if he is matched with an AI virtual object with a higher player confrontation mode, then This will make it more difficult for player A to operate, thereby reducing the player's interest.
  • an AI virtual object with a low confrontation mode is determined based on the specific attack attribute of player A, but player A prefers to dress up the virtual object without considering that player A prefers to dress up the virtual object, which leads to the problem of low flexibility of the AI operation mode.
  • This embodiment can adjust the AI operation mode to an operation mode with a more beautiful dress-up skin and a lower confrontation attribute by obtaining the operation information that player A likes to dress up the virtual object, thereby achieving the purpose of flexibly adjusting the AI operation mode.
  • the second operation information can be understood as operation information with a low similarity to the first operation information, that is, different from the first operation information
  • the second operation mode can be understood as an operation mode corresponding to the second operation information.
  • the operation information may include, but is not limited to, the number of operation instructions of the virtual object controlled by the user of the cloud game in each period of time during the game process, whether the high-level operation is completed, and the operation speed of executing the operation instruction completed at one time, such as the number of card draws, the number of challenges, the number of purchases, and other information.
  • player A and player B jointly control the first virtual object to complete a game, as shown in (a) of FIG3 .
  • Player A controls the first virtual object 302 to complete the first 3 minutes of the game process.
  • the rest of the game process is completed by player B controlling the first virtual object 302. Since player A's game operation level is relatively poor, after obtaining the first operation information generated by player A on the first virtual object 302 within the first 3 minutes of the game process, the game background learns that player A has not released a set of combos, so the first operation mode 304 corresponding to the second virtual object 306 controlled by artificial intelligence is determined as a novice operation mode, as shown in (b) of FIG3 .
  • player B controls the first virtual object 302 to complete the remaining game process. Since player B has a high level of game operation skills, the first operation mode 304 corresponding to the second virtual object 306 is adjusted to the second operation mode 308 according to the second operation information that player B has completed multiple difficult operations in a short period of time. That is, the novice operation mode is adjusted to the expert operation mode, thereby achieving the purpose of adjusting the operation mode of artificial intelligence according to the real-time operation level of the user, and further realizing the technical effect of improving the flexibility of virtual object control.
  • This embodiment adjusts the operation mode corresponding to the second virtual object in real time by acquiring the real-time operation information of the game player in the game, and the cloud game process runs on the server side, which can directly read the image and operation instruction stream from the video memory, and the intermediate process does not need to be landed, which greatly reduces the delay, and can achieve real-time acquisition and real-time adjustment, and achieves the technical effect of improving the efficiency of virtual object control.
  • shooting accuracy is the main gameplay of the game
  • shooting accuracy is the main gameplay of the game
  • virtual objects and virtual props in the game there is also a gameplay in which players collect costumes or skins.
  • a second virtual object 406 of an average level is pre-matched based on the player's level.
  • the first operation mode 404 corresponding to the second virtual object 406 is the shooting mode. Since player C is a player who likes to collect costumes but has a low shooting level, the operation mode corresponding to the second virtual object 406 can be adjusted by obtaining the costume attributes of the first virtual object 402 controlled by player C.
  • the costume attributes may include, but are not limited to, the number of skins owned, the rarity of the skin, and the like.
  • the first operation mode corresponding to the second virtual object 406 can be adjusted.
  • 404 e.g., normal dressing mode
  • a second operating mode 408 is adjusted to a second operating mode 408 with a higher aesthetics (e.g., reaching a preset aesthetics), thereby attracting players to complete the shooting game, achieving the purpose of increasing players' interest, and further achieving the technical effect of increasing the diversity of virtual object control.
  • real-time collection of operation information can be achieved through the cloud gaming scene, and the operation mode of the virtual object controlled by artificial intelligence can be determined based on the operation information.
  • real-time adjustment of the operation mode of the manually controlled virtual object can be achieved, thereby achieving the purpose of updating the game operation mode of the artificial intelligence according to the user's real-time game operation level, thereby achieving the technical effect of improving the control flexibility of the virtual object, and thus solving the technical problem of low control flexibility of the virtual object.
  • the method further includes:
  • Second operation information is determined.
  • the starting time point may be, but is not limited to, a specific moment in the running process of a cloud game.
  • This embodiment may be, but is not limited to, understood as after obtaining the first operation information generated by the user for the first virtual object, at this time, based on the first operation information, determining that the operation mode corresponding to the second virtual object is the first operation mode, further determining the start time of the first operation mode, starting from the start time of the first operation mode, and then obtaining multiple operation information of the user for the first virtual object during the running process of the cloud game, and determining the second operation information based on the operation information generated within a period of time after the start time of the first operation mode.
  • the second operation mode is determined based on the operation information of the players within a period of time, rather than through the operation information of a whole game.
  • the reason is that there may be a large gap in the levels of the players in different time periods during a whole game. For example, the player's level may be poor at the beginning, but the player's level may be particularly high over a period of time. If simply taking the average experience value of the players, it will lead to the inability to accurately determine the operation mode based on the player's operation information, and thus there is a technical problem of low accuracy in determining the operation mode.
  • this embodiment after determining the first operation mode based on the operation information of the player in a period of time, the operation information in the next period of time is re-acquired to adjust the operation mode. Compared with taking the average experience value of the player based on the operation information of the entire game, this embodiment achieves the technical effect of improving the accuracy of determining the operation mode by dividing the time period.
  • the starting time point of the first operation mode is determined; multiple operation information of the user on the first virtual object after the starting time point and during the running of a cloud game is obtained; based on the multiple operation information, the second operation information is determined, thereby achieving the purpose of collecting operation information by dividing the time period, thereby achieving the technical effect of improving the accuracy of determining the operation mode.
  • determining the second operation information based on the multiple operation information includes:
  • the first target operation information is determined as the second operation information.
  • the operation information can be understood as a collection of multiple operations performed by the user on the first virtual object during the operation of a cloud game.
  • the operation information can be, but is not limited to, key operation information and regular operation information.
  • the key operation information is multiple operation information with a certain degree of difficulty, importance, or rarity.
  • the key operation information can also be operation information formed by combining multiple regular operation information in a certain way. No unnecessary limitations are made here.
  • the identified multiple operation information may be compared with preset key operation information for information similarity, but is not limited to this, and the operation information whose similarity with the key operation information is greater than or equal to a first preset threshold is determined as the first target operation information.
  • the key operation may also be triggered by the user accidentally, it is also necessary to obtain the number of first target operation information, and determine the information whose number of first target operation information is greater than or equal to a second preset threshold as the second operation information.
  • the key operation information is operation information with a certain degree of difficulty or rarity
  • the user's ability to complete the key operation also indirectly proves that the user has a certain gaming experience or a high gaming level.
  • the problem of low accuracy For example, the user accidentally triggers the key operation, but in fact the user's overall game level is low. If the operation mode corresponding to the second virtual object is adjusted to a high-difficulty operation mode, it will be detrimental to the user's gaming experience.
  • this embodiment takes into account the combination of the number of key operation information, and thus adopts a technical means of combining the key operation with the number of operations to achieve the technical effect of accurately determining the second operation information.
  • the first virtual object 502 controlled by the player continuously releases a high-difficulty skill three times within 3 minutes, and causes 300 HP of damage to the second virtual object 504 controlled by the artificial intelligence, reaching 80% of the damage.
  • the collected operation information corresponding to the damage value, operation instructions, etc. is input into the model, and the result is that the player's level is relatively high at this time (for example, reaching a preset level), so the operation mode with a lower difficulty level corresponding to the second virtual object 504 is adjusted to an operation mode with a higher difficulty level.
  • the level of difficulty can be determined by a preset difficulty threshold. If the difficulty is lower than the difficulty threshold, the difficulty level is relatively low, and vice versa.
  • determining the second operation information based on the multiple operation information includes:
  • the multiple operation information are input into a target model, wherein the target model is a neural network model trained by using multiple sample operation information and used to identify the user's operation information on the first virtual object.
  • the second target operation information is determined as the second operation information.
  • the second target operation information is the operation information output by the target model.
  • the second target operation information is compared with the first operation information for similarity. If the similarity is greater than or equal to a certain preset threshold, the second target operation information is determined as the second operation information.
  • multiple levels can be set according to the difficulty of the game, the gameplay, etc., and the comprehensive data of the second target operation information and the first operation information are calculated respectively. If the levels indicated by the comprehensive data of the second target operation information and the first operation information are different, the second target operation information is determined as the second operation information.
  • This embodiment uses a trained neural network model to determine the operation information, thereby improving the accuracy of determining the operation information.
  • the method before inputting the plurality of operation information into the target model, the method comprises:
  • each sample operation information includes a current environment parameter, a current behavior parameter, and a current sample result
  • the current environment parameter being a parameter related to the environment in which the operation corresponding to the sample operation information is executed
  • the current behavior parameter being a behavior type corresponding to the operation corresponding to the sample operation information
  • the current sample result being information corresponding to the operation performed by the second virtual object that matches the operation corresponding to the sample operation information
  • the next sample is obtained from the multiple sample operation information, and the next sample is determined as the current sample.
  • the intermediate model is a model when the neural network model is not trained.
  • the training process of the target model may include but is not limited to being understood as collecting the environmental parameters, action parameters, and result parameters of the players during the entire game process of the battle between players.
  • the environmental parameters may be understood as, but are not limited to, parameters related to the environment in which the player's operation is executed during the game process, and may include, but are not limited to, the position information of the player when receiving an attack, the movement information of the player when releasing a skill, the environmental information around the player when releasing a skill, etc., and no unnecessary restrictions are made here.
  • the behavioral parameters may be understood as, but are not limited to, the behavioral parameters corresponding to the player's operation during the game process, and may include, but are not limited to, specific operation instructions for releasing a difficult skill once, operation instructions to respond to when receiving other operation instructions, and so on.
  • the current sample result is information corresponding to the operation corresponding to the sample operation information and performed by the second operation object. In this embodiment, the current sample result may also refer to the operation result when the player fights against the player.
  • the current sample is input into the intermediate model to obtain the intermediate operation information output by the intermediate model, and the information similarity between the intermediate operation information and the current sample result is compared.
  • the information similarity is greater than or equal to the third preset threshold, it is determined that the intermediate model has reached the convergence condition, and the intermediate model is determined as the target model; when the information similarity is less than the third preset threshold, it is determined that the intermediate model has not reached the convergence condition, and other samples are obtained from multiple sample operation information, and the above steps are performed on other sample information to perform model training until the intermediate model reaches the convergence condition.
  • the performance of the target model can be measured by using, but is not limited to, a loss function.
  • the distance from the "incentive value" generated by the model's predicted action is used as the loss function, for example, as shown in the following formula (1):
  • L is the abbreviation of LOSE, which means loss value.
  • the calculation process of the loss function is actually to calculate a mean square error, where the max a′ Q(s′,a′) function represents the target value. For example, in a confrontation game, the virtual object can lose 1 or 2 drops of health, but if you want to make the model perform better, you will choose a maximum loss value, which means the maximum damage value inflicted by the first virtual object controlled by the user.
  • the output of the function represents the value of the second virtual object controlled by the artificial intelligence. The difference between the two values is squared and divided by 2 to get the loss value L.
  • the purpose of model training is to make the loss value L equal to 0 or close to 0.
  • L is 0, it means that the model has reached a perfect state, then the model training is completed and the model is determined as the target model.
  • the distance between the actual incentive value and the incentive value generated by the model's predicted action is used as the loss function, and the model parameters are updated through back propagation to generate the final target model.
  • the algorithm that uses nonlinear propagation to approximate the Q value is unstable, and the model is difficult to converge in many cases, so the use of experience replay can make the model converge quickly.
  • online collection of operation information such as images and operation instructions will generate a large bandwidth burden, thereby affecting the player's gaming experience.
  • the cloud game itself runs on the server side, and directly using the resources of the server side to collect operation information will not generate additional bandwidth overhead, and the collection cost is low.
  • the cloud game service cluster itself has sufficient GPU computing power, and the computing power resources of the server can be fully utilized during the model training process to reduce computing costs.
  • the method further comprises:
  • the image information recognition module in the target model is used to perform image recognition on the image information to obtain processed operation information;
  • the processed operation information is input into the operation information recognition module in the target model to obtain second target operation information.
  • the purpose of obtaining the second target operation information by using image recognition and model training can be achieved, thereby realizing the technical effect of improving the accuracy of obtaining the operation information.
  • the advantage of low image acquisition cost when cloud games are running on the server is utilized to collect the player's real-time game screen, and the artificial intelligence operation instructions or operation modes that are closest to the player's actual operation level are obtained through model calculation.
  • model training and image acquisition are completed in the cloud, so there will be no additional delay due to data interaction or command issuance.
  • the opponent's response will be more timely, the game experience will be better, and the model parameters can be updated in real time online.
  • obtaining first operation information generated by a user on a first virtual object during the running of the cloud game, determining a first operation mode corresponding to a second virtual object based on the first operation information, and controlling the second virtual object according to the first operation mode includes:
  • the first operation information generated by the user on the first virtual object within a first time period during the operation of the cloud game is obtained, and a first operation mode is determined based on the first operation information, and the behavior operation of the second virtual object after the first time period is controlled according to the first operation mode.
  • the behavioral operation of controlling the second virtual object according to the first operation mode can be understood, but is not limited to, as determining that the operation level corresponding to the first virtual object is a low difficulty level (for example, lower than a preset difficulty threshold) based on the operation information generated by the user for the first virtual object, then setting the operation mode corresponding to the second virtual object to low difficulty, thereby controlling the second virtual object to execute medium or low difficulty operation instructions after the first time period.
  • a low difficulty level for example, lower than a preset difficulty threshold
  • the purpose of determining the operation mode based on the operation information within a time period can be achieved, thereby realizing the technical effect of increasing the diversity of the operation modes.
  • the method further comprises:
  • Acquire second operation information generated by the user on the first virtual object in a second time period after the first time period during the running of the cloud game determine a second operation mode based on the second operation information, and control the behavior operation of the second virtual object after the second time period according to the second operation mode.
  • the purpose of flexibly controlling the second virtual object can be achieved, and the technical effect of improving the flexibility of controlling the virtual object is realized.
  • controlling the second virtual object according to the first operation mode includes:
  • Controlling the second virtual object according to the second operation mode includes: controlling the second virtual object to execute at least one second operation instruction corresponding to the second operation mode.
  • the operation instruction executed by the second virtual object is adjusted to an operation instruction that is more difficult to release, thereby achieving the purpose of flexibly adjusting the specific operation instructions of the second virtual object according to the player's operation information on the first virtual object, and achieving the technical effect of improving the flexibility of virtual object control.
  • the above virtual object control method may be applied in a specific application scenario. As shown in FIG6 , the method may include the following steps S602-S614:
  • Step S602 detecting whether the player has entered the game connection.
  • Step S604 matching opposing players based on basic information such as the player's level, experience, winning rate, etc.
  • Step S606 if the match is successful, the game scene of real-person battle is entered, that is, the first virtual object controlled by the player fights against the third virtual object controlled by the matched opponent player, and the game is completed.
  • Step S608 when the first virtual object controlled by the player plays a game with the third virtual object controlled by the opposing player, the player's behavior and other operation information are periodically collected using real-time image information and action results, for example, approximately once every 32ms, and values are captured from the real-time image, such as blood value, real-time image information including the current player's position and the player's behavior and operation, and action results may include but are not limited to blood value deduction in battle, etc.
  • Step S610 After collecting the operation information, the operation information is input into the deep learning network model for training.
  • the construction of the Deep Q Network model can be, but is not limited to, as shown in FIG7 .
  • the state parameter 702 (state) indicates the current environment of the player, such as a two-player game scene, the position of the target of the first virtual object, etc.
  • the behavior parameter 704 (action) indicates the behavior operation taken by the first virtual object on the target, such as moving forward or kicking.
  • Neural networks are the basis of deep learning networks.
  • the neural network model shown in Figure 8 includes an input layer, three convolutional layers, two fully connected layers, and an output layer.
  • the purpose of the convolutional layer is to extract information from the image and perform image recognition, such as identifying where the virtual object is located in the image, or to capture some information from the image, such as the maximum winning rate of the virtual object controlled by the player, etc.
  • the purpose of training the neural network model is to determine that when the first virtual object is in a certain state, the second virtual object simulated by artificial intelligence can adopt a strategy corresponding to the state.
  • the training of the neural network model requires setting an optimization goal so that the neural network model can output a value that meets the requirements through training.
  • the essence of building a neural network model is to build a formula to adjust the parameters of the neural network model, such as using the loss function shown in formula (1) above to measure the performance of the neural network model.
  • Steps S612-S614 if the match is not successful, then enter the game scene of artificial intelligence battle, collect the real-time picture of the first virtual object controlled by the player, here each frame of the image can be collected, and the collected data is input into the trained neural network model for calculation, to obtain the best response of the second virtual object controlled by artificial intelligence, such as artificial intelligence operation instructions or operation mode, etc.
  • artificial intelligence such as artificial intelligence operation instructions or operation mode, etc.
  • model training and image collection are completed in the cloud, so there will be no additional delay caused by data interaction or instruction issuance.
  • the opponent's response will be more timely, the game experience will be better, and the model parameters can be updated online in real time without occupying bandwidth to be sent to the local client.
  • the flow of the model training process can be shown as the training sequence diagram of Figure 9.
  • the user client 902 uploads the operation instructions input by the player to the cloud game sandbox process 904 in the server, and the data acquisition module 906 captures the current image and the player's operation instructions, as well as the blood deduction information after the operation from the game sandbox process 904, completes the collection of images, operation instructions, and incentive values, and then inputs the images and operation instructions as input data into the Deep Q Network model 908, trains the Deep Q Network model 908, and uses back propagation to update the model parameters.
  • the process of the battle between the player and the artificial intelligence can be shown in the timing diagram of Figure 10.
  • the user client 1002 inputs the player's operation instruction into the cloud game sandbox process 1004, and the data acquisition module 1006 performs image acquisition from the cloud game sandbox process 1004, extracts the player's current game screen and operation instruction, and inputs the extracted game screen and operation instruction into the trained Deep Q Network model 1008.
  • the Deep Q Network model 1008 returns the operation instruction of the artificial intelligence to the cloud game sandbox process 1004.
  • the cloud game sandbox process 1004 receives the returned operation instruction and inputs the operation instruction into the game process, so that the artificial intelligence operates according to the operation instruction, and then returns the game battle screen to the cloud game server, and the cloud game server returns it to the user client 1002 for display.
  • an AI that is more in line with human thinking can be generated, and the unreasonable parts of AI can be optimized, so that players feel that AI is alive and full of vitality like real people, thereby better attracting players and enhancing player stickiness.
  • the matching mechanism of AI can also be optimized, and the game level of players can be more accurately evaluated, so as to match opponents and teammates of similar levels, ensure that the game difficulty is moderate, and allow players to obtain a better game experience and enjoy the fun of competition.
  • a virtual object control device for implementing the above virtual object control method is also provided. As shown in FIG11 , the device includes:
  • the first display unit 1102 is used to display a first virtual object and a second virtual object participating in a cloud game during the running of the cloud game, wherein the first virtual object is a virtual object controlled by a user of the cloud game, and the second virtual object is a virtual object controlled by artificial intelligence;
  • the first determining unit 1104 is used to obtain first operation information generated by the user on the first virtual object during the operation of the cloud game, determine a first operation mode corresponding to the second virtual object based on the first operation information, and control the second virtual object according to the first operation mode;
  • the first adjustment unit 1106 is used to adjust the first operation mode corresponding to the second virtual object to the second operation mode based on the second operation information generated by the user for the first virtual object, and control the second virtual object according to the second operation mode, wherein the first operation information is different from the second operation information, and the first operation mode is different from the second operation mode.
  • the apparatus further comprises:
  • a second determining unit configured to determine a starting time point of the first operation mode after acquiring the first operation information
  • a first acquisition unit is used to acquire a plurality of operation information of the user on the first virtual object after the starting time point and during the running process of the cloud game;
  • the second acquisition unit is used to determine second operation information based on the multiple operation information.
  • the second acquiring unit includes:
  • a first determination module used to determine, from the plurality of operation information, first target operation information whose information similarity with the key operation information is greater than or equal to a first preset threshold
  • the second determining module is used to determine the first target operation information as the second operation information when the amount of the first target operation information is greater than or equal to a second preset threshold.
  • the second acquiring unit further includes:
  • the first input module is used to input multiple operation information into a target model, wherein the target model is a neural network model trained using multiple sample operation information and used to identify the user's operation information on the first virtual object.
  • a second acquisition module used to acquire second target operation information output by the target model
  • the third determining module is used to determine the second target operation information as the second operation information when the second target operation information is different from the first operation information.
  • the apparatus further comprises:
  • the fourth determination module includes a first acquisition submodule, a first input submodule, a first determination submodule and a second determination submodule, and is used to perform the following steps before inputting a plurality of operation information into the target model until the target model is obtained.
  • the first acquisition submodule is used to obtain the current sample from multiple sample operation information, wherein each sample operation information includes current environment parameters, current behavior parameters, and current sample results, the current environment parameters are parameters related to the environment when the operation corresponding to the sample operation information is executed, the current behavior parameters are the behavior type corresponding to the operation corresponding to the sample operation information, and the current sample result is information corresponding to the operation performed by the second virtual object that matches the operation corresponding to the sample operation information.
  • the first input submodule is used to input the current sample into the intermediate model to obtain the intermediate operation information output by the intermediate model, where the intermediate model is the model when the neural network model is not trained.
  • the first determination submodule is used to determine that the intermediate model has reached a convergence condition and determine the intermediate model as a target model when the information similarity between the intermediate operation information and the current sample result is greater than or equal to a third preset threshold.
  • the second determination submodule is used to determine that the intermediate model has not reached the convergence condition when the information similarity between the intermediate operation information and the current sample result is less than a third preset threshold, obtain the next sample from multiple sample operation information, and determine the next sample as the current sample.
  • the apparatus further comprises:
  • a first recognition module is used to, after inputting the multiple operation information into the target model, use the image information recognition module in the target model to perform image recognition on the image information in the case where the operation information is image information collected during the operation of a cloud game, so as to obtain processed operation information;
  • the second input module is used to input the processed operation information into the operation information recognition module in the target model to obtain second target operation information.
  • the first determining unit 1104 includes:
  • the third acquisition module is used to obtain first operation information generated by the user on the first virtual object within a first time period during the operation of the cloud game, determine a first operation mode based on the first operation information, and control the behavior operation of the second virtual object after the first time period according to the first operation mode.
  • the apparatus further comprises:
  • the fourth acquisition module is used to obtain second operation information generated by the user for the first virtual object in a second time period after the first time period during the running of the cloud game, after controlling the behavior operation of the second virtual object after the first time period according to the first operation mode, and determine the second operation mode based on the second operation information, and control the behavior operation of the second virtual object after the second time period according to the second operation mode.
  • the first determination unit 1104 includes a first control module, which is used to control the second virtual object to execute at least one first operation instruction corresponding to the first operation mode;
  • the first adjustment unit 1106 includes a second control module, which is used to control the second virtual object to execute at least one second operation instruction corresponding to the second operation mode.
  • the specific implementation of the device can refer to the above-mentioned embodiment of controlling the virtual object, which will not be described in detail here.
  • an electronic device for implementing the above-mentioned virtual object control method is also provided.
  • the electronic device includes a memory 1202 and a processor 1204 .
  • the memory 1202 stores a computer program
  • the processor 1204 is configured to execute the steps in any one of the above method embodiments by executing the computer program.
  • the electronic device may be located in at least one network device among a plurality of network devices of a computer network.
  • the structure shown in FIG. 12 is only an illustrative embodiment, and the electronic device may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, and other terminal devices.
  • FIG. 12 does not limit the structure of the above electronic device.
  • the electronic device may also include more or fewer components (such as a network interface, etc.) than those shown in FIG. 12, or have a configuration different from that shown in FIG. 12.
  • the memory 1202 may be used to store software programs and modules, such as program instructions and/or modules corresponding to the control method and device of the virtual object in the embodiment of the present application.
  • the processor 1204 executes various functional applications and data processing by running the software programs and modules stored in the memory 1202, that is, the control method of the virtual object described above is realized.
  • the memory 1202 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory.
  • the memory 1202 may further include a memory remotely arranged relative to the processor 1204, and these remote memories may be connected to the terminal device via a network.
  • the above-mentioned network examples include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.
  • the memory 1202 may be specifically used, but not limited to, for storing information such as the first operation information and the second operation information.
  • the above-mentioned memory 1202 may include, but is not limited to, the first display unit 1102, the first determination unit 1104, and the first adjustment unit 1106 in the control device of the above-mentioned virtual object.
  • other module units in the control device of the virtual object may also be included but not limited to the above, which will not be described in detail in this example.
  • the transmission device 1206 is used to receive or send data via a network.
  • the network may specifically include a wired network and a wireless network.
  • the transmission device 1206 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices and routers via a network cable, so as to communicate with the Internet or a local area network.
  • the transmission device 1206 is a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.
  • RF radio frequency
  • the electronic device further includes: a display 1208 for displaying the first operation information, the second operation information, etc.; and a connection bus 1210 for connecting various module components in the electronic device.
  • the terminal device or server may be a node in a distributed system, wherein the distributed system may be a blockchain system, and the blockchain system may be a distributed system formed by connecting multiple nodes through network communication.
  • a peer-to-peer (P2P) network may be formed between the nodes, and any form of computing device, such as a server, a terminal device or other electronic device, may become a node in the blockchain system by joining the peer-to-peer network.
  • P2P peer-to-peer
  • an embodiment of the present application includes a computer program product, which includes a computer program and/or instructions carried on a computer-readable medium, and the computer program and/or instructions contain program codes for executing the method shown in the flow chart.
  • the computer program can be downloaded and installed from a network through a communication part, and/or installed from a removable medium.
  • various functions defined in the system of the present application are executed.
  • a computer-readable storage medium for storing computer instructions.
  • a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in the above-mentioned various implementations.
  • the program can be stored in a computer-readable storage medium.
  • the storage medium can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.
  • the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above computer-readable storage medium.
  • the technical solution of the present application, or the part that contributes to the relevant technology, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to enable one or more computer devices (which can be personal computers, servers or network devices, etc.) to execute all or part of the steps of the methods of each embodiment of the present application.
  • the disclosed client can be implemented in other ways.
  • the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.
  • Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of units or modules can be electrical or other forms.
  • the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one device or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Processing Or Creating Images (AREA)

Abstract

一种虚拟对象的控制方法、装置和存储介质及电子设备。其中，该方法包括：在一局云游戏的运行过程中，显示参与一局云游戏的第一虚拟对象、以及第二虚拟对象；获取在该局云游戏的运行过程中用户针对第一虚拟对象所产出的第一操作信息，并基于第一操作信息确定第二虚拟对象对应的第一操作模式，以及按照第一操作模式控制第二虚拟对象；在获取到在该局云游戏的运行过程中用户针对第一虚拟对象所产出的第二操作信息的情况下，基于第二操作信息将第二虚拟对象对应的第一操作模式调整为第二操作模式，并按照第二操作模式控制第二虚拟对象。本方法解决了虚拟对象的控制的灵活度较低的技术问题。

Description

虚拟对象的控制方法、装置和存储介质及电子设备
本申请要求于2022年11月22日提交中国专利局、申请号为202211466604.6、名称为“虚拟对象的控制方法、装置和存储介质及电子设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及计算机领域，具体而言，涉及一种虚拟对象的控制方法、装置和存储介质及电子设备。
背景技术
在云游戏场景下，由用户(或称为游戏玩家)控制的虚拟对象与人工智能控制的虚拟对象进行交互，以达到获得经验、材料或者通关奖励的目的。在面对不同游戏等级的玩家时，相关技术通常通过简单地调整人工智能控制的虚拟对象的数值来调节人工智能强度，以避免人工智能强度与玩家水平差距过大的问题。
但是，当同一个玩家在游戏过程中表现出不同的游戏水平时，上述调节方式无法根据玩家的游戏水平灵活地调整人工智能控制的虚拟对象的数值。因此，相关技术中存在对虚拟对象的控制灵活度较低的问题。
针对上述的问题，目前尚未提出有效的解决方案。
发明内容
本申请实施例提供了一种虚拟对象的控制方法、装置和存储介质及电子设备，以至少解决虚拟对象的控制的效率较低的技术问题。
根据本申请实施例的一个方面，提供了一种虚拟对象的控制方法，包括：
在一局云游戏的运行过程中，显示参与所述一局云游戏的第一虚拟对象、以及第二虚拟对象，其中，所述第一虚拟对象为所述云游戏的用户所控制的虚拟对象，所述第二虚拟对象为人工智能控制的虚拟对象；
获取在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象所产出的第一操作信息，并基于所述第一操作信息确定所述第二虚拟对象对应的第一操作模式，以及按照所述第一操作模式控制所述第二虚拟对象；
在获取到在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象所产出的第二操作信息的情况下，基于所述第二操作信息将所述第二虚拟对象对应的所述第一操作模式调整为第二操作模式，并按照所述第二操作模式控制所述第二虚拟对象，其中，所述第一操作信息不同于所述第二操作信息，所述第一操作模式不同于所述第二操作模式。
根据本申请实施例的另一方面，还提供了一种虚拟对象的控制装置，包括：
第一显示单元，用于在一局云游戏的运行过程中，显示参与所述一局云游戏的第一虚拟对象、以及第二虚拟对象，其中，所述第一虚拟对象为所述云游戏的用户所控制的虚拟对象，所述第二虚拟对象为人工智能控制的虚拟对象；
第一确定单元，用于获取在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象所产出的第一操作信息，并基于所述第一操作信息确定所述第二虚拟对象对应的第一操作模式，以及按照所述第一操作模式控制所述第二虚拟对象；
第一调整单元，用于在获取到在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象所产出的第二操作信息的情况下，基于所述第二操作信息将所述第二虚拟对象对应的所述第一操作模式调整为第二操作模式，并按照所述第二操作模式控制所述第二虚拟对象，其中，所述第一操作信息不同于所述第二操作信息，所述第一操作模式不同于所述第二操作模式。
根据本申请实施例的又一个方面，提供一种计算机程序产品或计算机程序，该计算机程序产品或计算机程序包括计算机指令，该计算机指令存储在计算机可读存储介质中。计算机设备的处理器从计算机可读存储介质读取该计算机指令，并执行该计算机指令，使得该计算机设备执行如上所述的虚拟对象的控制方法。
根据本申请实施例的又一方面，还提供了一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其中，上述处理器通过计算机程序执行上述的虚拟对象的控制方法。
在本申请实施例中，利用云游戏实现对操作信息的实时采集，再基于采集到的操作信息确定人工智能控制的虚拟对象对应的操作模式，再基于上述操作信息在云游戏过程中的实时变化，灵活调整人工智能控制的虚拟对象对应的操作模式，进而达到了根据用户的实时游戏水平更新人工智能的操作模式的目的，从而实现了提高虚拟对象的控制灵活性的技术效果，进而解决了虚拟对象的控制灵活度较低的技术问题。
附图简要说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1是根据本申请实施例的一种虚拟对象的控制方法的应用环境的示意图；
图2是根据本申请实施例的一种虚拟对象的控制方法的流程的示意图；
图3是根据本申请实施例的一种虚拟对象的控制方法的示意图；
图4是根据本申请实施例的另一种虚拟对象的控制方法的示意图；
图5是根据本申请实施例的另一种虚拟对象的控制方法的示意图；
图6是根据本申请实施例的另一种虚拟对象的控制方法的示意图；
图7是根据本申请实施例的另一种虚拟对象的控制方法的示意图；
图8是根据本申请实施例的另一种虚拟对象的控制方法的示意图；
图9是根据本申请实施例的另一种虚拟对象的控制方法的示意图；
图10是根据本申请实施例的另一种虚拟对象的控制方法的示意图；
图11是根据本申请实施例的一种虚拟对象的控制装置的示意图；
图12是根据本申请实施例的一种电子设备的结构示意图。
具体实施方式
为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述。显然，所描述的实施例仅仅是本申请一部分的实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本申请保护的范围。
需要说明的是，本申请的说明书和权利要求书及附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、***、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
为方便理解，对本申请实施例中所涉及的术语进行解释：
人工智能(Artificial Intelligence，AI)是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能，感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用***。换句话说，人工智能是计算机科学的一个综合技术，它企图了解智能的实质，并生产出一种新的能以人类智能相似的方式做出反应的智能机器。人工智能也就是研究各种智能机器的设计原理与实现方法，使机器具有感知、推理与决策的功能。
人工智能技术是一门综合学科，涉及领域广泛，既有硬件层面的技术也有软件层面的技术。人工智能基础技术一般包括如传感器、专用人工智能芯片、云计算、分布式存储、大数据处理技术、操作/交互***、机电一体化等技术。人工智能软件技术主要包括计算机视觉技术、语音处理技术、自然语言处理技术以及机器学习/深度学习等几大方向。
随着人工智能技术研究和进步，人工智能技术在多个领域展开研究和应用，例如常见的智能家居、智能穿戴设备、虚拟助理、智能音箱、智能营销、无人驾驶、自动驾驶、无人机、机器人、智能医疗、智能客服等，相信随着技术的发展，人工智能技术将在更多的领域得到应用，并发挥越来越重要的价值。
云游戏(Cloud gaming)又可称为游戏点播(gaming on demand)，是一种以云计算技术为基础的在线游戏技术。云游戏技术使图形处理与数据运算能力相对有限的轻端设备(thin client)能运行高品质游戏。在云游戏场景下，游戏并不在玩家游戏终端，而是在云端服务器中运行，并由云端服务器将游戏场景渲染为视频音频流，通过网络传输给玩家游戏终端。玩家游戏终端无需拥有强大的图形运算与数据处理能力，仅需具有基本的流媒体播放能力与获取玩家输入指令并发送给云端服务器的能力即可。
本申请实施例提供的方案涉及人工智能的图像识别、云游戏等技术，具体通过如下实施例进行说明。
根据本申请实施例的一个方面，提供了一种虚拟对象的控制方法。作为一种实施方式，上述虚拟对象的控制方法可以但不限于应用于如图1所示的环境中。其中，可以但不限于包括用户设备102以及服务器112，该用户设备102上可以但不限于包括显示器104、处理器106及存储器108，该服务器112包括数据库114以及处理引擎116。
具体的控制过程可包括如下步骤：
步骤S102，用户设备102获取来自于第一虚拟对象1002对应的客户端中的用户针对第一虚拟对象的第一操作信息；
步骤S104-S106，通过网络110将用户针对第一虚拟对象的第一操作信息发送至服务器112；
步骤S108，服务器112通过处理引擎基于第一操作信息确定第二虚拟对象1004对应的第一操作模式；
步骤S110-S112，通过网络110将第二虚拟对象1004对应的第一操作模式的第二操作信息发送至用户设备102，用户设备102通过处理器106处理第一操作模式的第二操作信息，并将基于第一操作模式的第二操作信息控制第二虚拟对象1004释放技能的过程显示在客户端，以及将第一操作信息和第二操作信息存储在存储器104。
步骤S114，用户设备102向第二虚拟对象1004所在的设备发送位置信息的提示标识。
除图1示出的示例之外，上述步骤可以由客户端或服务器独立完成，或由客户端和服务器共同协作完成，如由用户设备102执行上述S108等步骤，从而减轻服务器112的处理压力。该用户设备102包括但不限于手持设备(如手机)、笔记本电脑、台式电脑、车载设备等，本申请并不限制用户设备102的具体实现方式。
作为一种实施方式，如图2所示，虚拟对象的控制方法包括步骤S202-S206：
S202，在一局云游戏的运行过程中，显示参与该局云游戏的第一虚拟对象、以及第二虚拟对象，其中，第一虚拟对象为云游戏的用户所控制的虚拟对象，第二虚拟对象为人工智能的虚拟对象；
S204，获取在该局云游戏的运行过程中用户针对第一虚拟对象所产出的第一操作信息，并基于第一操作信息确定第二虚拟对象对应的第一操作模式，以及按照第一操作模式控制第二虚拟对象；
S206，在获取到在该局云游戏的运行过程中用户针对第一虚拟对象所产出的第二操作信息的情况下，基于第二操作信息将第二虚拟对象对应的第一操作模式调整为第二操作模式，并按照第二操作模式控制第二虚拟对象，其中，第一操作信息不同于第二操作信息，第一操作模式不同于第二操作模式。
在本实施例中，上述虚拟对象的控制方法可以但不限于应用在云游戏的场景中。云游戏可以但不限于理解为，游戏玩家通过终端设备输入指令，而游戏动画效果的实时渲染、图形运算和数据处理由云服务器直接负责，这大大降低了对游戏玩家的终端设备的运算要 求。对传统游戏而言，这部分工作通常由游戏玩家的终端设备的主机来负责，所需要的大量算力需要体积较大且费用配置较高的终端设备来执行。在云游戏模式下，由于所有的图形运算、游戏场景渲染都从本地硬件中剥离出来，游戏玩家的终端设备只需负责显示和编码功能即可，不需要较高的功耗和存储空间。在相关技术中，在游戏玩家控制的虚拟对象与AI控制的虚拟对象进行交互的过程中，通常通过简单的调整AI控制的虚拟对象的数值来调节AI强度，调节后的AI强度过高或过低都会引起游戏玩家的反感，从而大大降低了游戏乐趣，故相关技术存在AI强度调整灵活性较低的技术问题。
在本实施例中，第一虚拟对象可以但不限于理解为由用户(即当前游戏玩家)控制的虚拟对象，例如用户可以控制虚拟对象的移动、挑战、释放技能等，这里不做多余限定。第二虚拟对象可以但不限于理解为由人工智能模拟用户控制的虚拟对象，这里的用户是指人类，即人工智能模拟人类的思维(具体为游戏玩家的思维)来控制虚拟对象。其中第一虚拟对象和第二虚拟对象之间的关系可以但不限于属于游戏的同一阵营或者敌对阵营，例如由用户控制的第一虚拟对象与人工智能模拟用户控制的第二虚拟对象为不同阵营的敌对关系。游戏后台可以基于用户的经验等级、历史胜率、操作评分等等因素综合考量用户的游戏水平，并基于用户的游戏水平使人工智能控制第二虚拟对象执行相应的操作，从而实现了根据游戏玩家的综合游戏水平确定人工智能模拟的操作模式的目的，达到了提高操作模式确定的准确性的技术效果。
在本实施例中，可以但不限于基于用户的历史信息和经验等级等预先确定出第二虚拟对象的游戏模式。在游戏过程中，实时获取由用户针对第一虚拟对象产出的第一操作信息，例如实时获取用户是否完成难度较高(例如达到预设的难度阈值)的操作指令、用户的操作速度、用户的挑战次数、用户的被击败次数等等。基于第一操作信息确定人工智能控制的第二虚拟对象对应的操作模式(也称为人工智能的操作模式)，例如用户在游戏开始时就完成了一次高难度操作，则根据第一操作信息将人工智能的等级调整至高难度等级，该难度等级与用户操作的难度相匹配，从而实现了根据用户的第一操作信息，确定人工智能的操作模式的目的，进而生成难度适中的AI，给玩家带来较好的游戏体验。
在本实施例中，操作模式可以是但不限于由难易程度确定的操作模式，例如游戏中的新手模式、普通模式和挑战模式等等，也可以是由游戏版块的功能目的确定的操作模式，例如游戏中的娱乐模式、正式模式等等，也可以是根据游戏玩家的玩法确定的不同的操作模式，比如在包含有收集、对抗等玩法的游戏中，玩家更倾向于收集玩法，那么可以根据玩家的玩法调整AI的操作模式为收集等级较高(例如，达到预设等级)的AI。例如射击类游戏中，用户喜欢收集各种各样道具的皮肤，而对抗射击水平较差，人工智能可以获取用户的游戏玩法，将其控制的虚拟对象对应的操作模式调整为装扮特效较好但是对抗水平较差的操作模式。
需要说明的是，相关技术通常根据玩家的经验数值确定AI的操作模式，并未考虑到玩家的游戏玩法。例如，玩家A更喜欢对虚拟对象进行装扮，而不喜欢操作虚拟对象进行攻击，即便玩家A的整体经验数值较高，若匹配给玩家对抗模式较高的AI虚拟对象，则 会导致玩家A操作难度较大，从而降低玩家的兴趣。或者，根据玩家A的具体攻击属性确定对抗模式较低的AI虚拟对象，但是并未考虑玩家A更喜欢对虚拟对象进行装扮，则导致AI的操作模式灵活性较低的问题。本实施例可以通过获取玩家A喜欢对虚拟对象进行装扮的操作信息，将AI的操作模式调整为拥有较美观的装扮皮肤且对抗属性较低的操作模式，从而实现灵活调整AI的操作模式的目的。
在本实施例中，第二操作信息可以理解为与第一操作信息相似度较低的操作信息，即不同于第一操作信息，第二操作模式可以理解为与第二操作信息对应的操作模式。操作信息可以但不限于包括云游戏的用户控制的虚拟对象在游戏进程中每一段时间内的操作指令的数量、是否完成高等级操作、一次完成的执行操作指令的操作速度，例如抽卡次数、挑战次数、购买次数等等信息。
进一步举例说明，玩家A和玩家B共同控制第一虚拟对象完成一局游戏，例如图3中的(a)所示。玩家A在游戏进程中的前3分钟控制第一虚拟对象302完成，在玩家A完成前3分钟的游戏进程后，其余游戏进程由玩家B控制第一虚拟对象302完成。由于玩家A的游戏操作水平较差，在获取到玩家A在游戏进程的前3分钟内针对第一虚拟对象302所产生的第一操作信息后，游戏后台获知玩家A并未释放出一套连招，故将由人工智能控制的第二虚拟对象306对应的第一操作模式304确定为新手操作模式，例如图3中的(b)所示。三分钟后，由玩家B控制第一虚拟对象302完成剩余的游戏进程，由于玩家B的游戏操作水平较高，那么根据玩家B短时间内连续完成多个高难度操作的第二操作信息，故将第二虚拟对象306对应的第一操作模式304调整为第二操作模式308，即将新手操作模式调整为高手操作模式，从而达到了根据用户的实时操作水平调整人工智能的操作模式的目的，进而实现了提高虚拟对象控制的灵活性的技术效果。
需要说明的是，相关技术中针对游戏玩家的操作水平对AI的操作模式的调整都发生在游戏开始前，但是忽略了游戏过程中针对同一虚拟对象游戏玩家的操作水平也可能出现较大的变化，因此相关技术存在对虚拟对象的控制不灵活的问题。本实施例通过游戏内游戏玩家的实时操作信息的获取，实时对第二虚拟对象对应的操作模式进行调整，且云游戏的进程在服务端运行，可以直接从显存读取图像和操作指令流，中间过程无需落地，极大的降低了延时性，能够做到实时获取和实时调整，实现了提高虚拟对象控制的效率的技术效果。
进一步举例说明，在射击类游戏中，虽然射击的准确度是游戏的主要玩法，但是由于游戏内虚拟对象的装扮和虚拟道具的装扮较多，也存在玩家收集装扮或者皮肤的玩法。例如图4中的(a)所示，预先基于玩家的等级匹配一个平均水平的第二虚拟对象406，此时第二虚拟对象406对应的第一操作模式404为射击模式。由于玩家C是一个喜欢收集装扮、但是射击水平较低的玩家，因此可以通过获取玩家C控制的第一虚拟对象402身上的装扮属性调整第二虚拟对象406对应的操作模式。装扮属性可以但不限于包括皮肤拥有的数量、皮肤的稀有度等等。通过获取第一虚拟对象402的装扮为稀有度较高(例如，达到预设的稀有度)的装扮，例如图4中(b)所示，可以将第二虚拟对象406对应的第一操作模式 404(例如普通装扮模式)调整为美观度较高(例如，达到预设的美观度)的第二操作模式408，从而吸引玩家完成射击游戏，达到提高玩家兴趣的目的，进而实现了提高虚拟对象控制的多样性的技术效果。
在本申请实施例中，通过云游戏场景可以实现对操作信息的实时采集，基于操作信息确定人工智能控制的虚拟对象的操作模式，且基于游戏过程中操作信息的变更，能够实现人工控制的虚拟对象的操作模式的实时调整，进而达到了根据用户的实时游戏操作水平更新人工智能的游戏操作模式的目的，从而实现了提高虚拟对象的控制灵活性的技术效果，进而解决了虚拟对象的控制灵活度较低的技术问题。
在实施例中，在获取用户针对第一虚拟对象所产出的第一操作信息之后，该方法还包括：
确定第一操作模式的起始时间点；
获取在起始时间点之后、在该局云游戏的运行过程中用户针对第一虚拟对象的多个操作信息；
基于该多个操作信息，确定第二操作信息。
在本实施例中，起始时间点可以但不限于为一局云游戏运行过程中的某具体时刻。本实施例可以但不限于理解为获取用户针对第一虚拟对象所产出的第一操作信息之后，此时基于第一操作信息确定第二虚拟对象对应的操作模式为第一操作模式，进一步确定第一操作模式的开始时刻，从第一操作模式的开始时刻开始，再获取该局云游戏运行过程中用户针对第一虚拟对象的多个操作信息，并基于第一操作模式的开始时刻后的一段时间内产出的操作信息确定第二操作信息。
需要说明的是，本实施例中，基于一段时间内的玩家的操作信息确定第二操作模式，而不是通过一整局的操作信息确定第二操作模式，理由在于一整局游戏进程中不同时间段内的玩家水平可能会存在较大差距，例如可能存在开始时玩家水平较差，而在一段时间内玩家水平特别高的情况，若单纯的取玩家的平均经验数值，则会导致无法准确的根据玩家的操作信息确定操作模式，进而存在操作模式的确定准确性较低的技术问题。
在本实施例中，通过在一段时间内的玩家的操作信息确定第一操作模式之后，再重新获取下一段时间内的操作信息，从而进行操作模式的调整。相比于根据整局游戏的操作信息取玩家的平均经验数值，本实施例通过时间段划分，实现了提高操作模式确定的准确性的技术效果。
通过本申请提供的实施例，确定第一操作模式的起始时间点；获取在起始时间点之后、在一局云游戏的运行过程中用户针对第一虚拟对象的多个操作信息；基于该多个操作信息，确定第二操作信息，达到了通过时间段划分进行操作信息的采集的目的，从而实现了提高操作模式确定的准确性的技术效果。
在实施例中，基于该多个操作信息，确定第二操作信息，包括：
从多个操作信息中确定出与关键操作信息之间信息相似度大于或等于第一预设阈值的第一目标操作信息；
在第一目标操作信息的数量大于或等于第二预设阈值的情况下，将第一目标操作信息确定为第二操作信息。
在本实施例中，操作信息可以理解为在一局云游戏的运行过程中用户针对第一虚拟对象产出的多个操作的集合，操作信息可以但不限于由关键操作信息和常规操作信息组成，关键操作信息为多个具有一定难度或者重要程度、稀有程度的操作信息，其中关键操作信息也可以是由多个常规操作信息通过一定方式组合形成的操作信息，这里不做多余限定。
在本实施例中，可以但不限于将识别到的多个操作信息与预设的关键操作信息进行信息相似度的比对，并将操作信息中与关键操作信息之间的相似度大于或者等于第一预设阈值的操作信息确定为第一目标操作信息。考虑到关键操作也可能是由用户偶然触发的，故还需要获取第一目标操作信息的数量，将第一目标操作信息的数量大于或者等于第二预设阈值的信息确定为第二操作信息。
需要说明的是，考虑到关键操作信息为具有一定难度或者稀有程度的操作信息，故用户能够完成关键操作也间接证明了用户具有一定的游戏经验或者较高的游戏水平。但是，若仅仅考虑到将是否完成关键操作作为判断AI的操作模式的依据，则会导致准确性不高的问题。例如，用户偶然触发关键操作，但是实际上用户的综合游戏水平较低，若调整第二虚拟对象对应的操作模式为高难度操作模式则不利于用户游戏的体验。进一步地，本实施例考虑到结合关键操作信息的数量，从而采用将关键操作与操作数量相结合的技术手段实现了准确的确定第二操作信息的技术效果。
进一步举例说明，例如图5所示，识别到由玩家控制的第一虚拟对象502在3分钟内连续释放三次高难度技能，并对人工智能控制的第二虚拟对象504造成300血量的伤害，达到伤害的百分之八十。将采集到伤害数值、操作指令等对应的操作信息输入到模型中，得到结果为此时玩家水平较高(例如，达到预设等级)，故将第二虚拟对象504对应的难度水平较低的操作模式调整为难度水平较高的操作模式。在实施例中，难度水平的高低可以通过预设的难度阈值来确定，难度低于难度阈值则为难度水平较低，反之则较高。
在实施例中，基于该多个操作信息，确定第二操作信息，包括：
将该多个操作信息输入目标模型，其中，目标模型为利用多个样本操作信息进行训练得到的、用于识别用户针对第一虚拟对象的操作信息的神经网络模型。
获取目标模型输出的第二目标操作信息；
在第二目标操作信息不同于第一操作信息的情况下，将第二目标操作信息确定为第二操作信息。
这样，可以达到结合关键操作和操作数量的目的，实现了提高第二操作信息确定的准确性的技术效果。
在本实施例中，在获取多个操作信息之后，例如玩家的操作指令、击败次数等等，还可以但不限于抓取玩家的当前游戏画面，提取游戏画面内的玩家信息，例如时间信息、游戏任务完成度等等，结合从游戏画面中提取的环境信息和操作指令，使用归一化、数据整合等处理手段将操作信息输入到已经完成训练的神经网络模型中。
在本实施例中，第二目标操作信息为目标模型输出的操作信息。将第二目标操作信息与第一操作信息进行相似度的比对，若相似度大于或者等于一定预设阈值，则将第二目标操作信息确定为第二操作信息。或者，可以根据游戏难度、玩法等设置多个级别，分别计算第二目标操作信息和第一操作信息的综合数据，若第二目标操作信息和第一操作信息的综合数据指示的级别不同，那么将第二目标操作信息确定为第二操作信息。
需要说明的是，相关技术中在确定操作信息或者操作模式时都使用简单的调整参数，本实施例使用训练完成的神经网络模型确定操作信息，提高了操作信息确定的准确性。
在实施例中，在将多个操作信息输入目标模型之前，该方法包括：
执行以下步骤，直至得到目标模型：
从多个样本操作信息中获取当前样本，其中，每个样本操作信息包含有当前环境参数、当前行为参数、以及当前样本结果，当前环境参数为样本操作信息对应的操作被执行时所处的环境相关的参数，当前行为参数为样本操作信息对应的操作对应的行为类型，当前样本结果为与样本操作信息对应的操作匹配的、由第二虚拟对象执行的操作所对应的信息；
将当前样本输入中间模型，得到中间模型输出的中间操作信息；
在中间操作信息与当前样本结果之间的信息相似度大于或等于第三预设阈值的情况下，确定中间模型达到收敛条件，并将中间模型确定为目标模型；
在中间操作信息与当前样本结果之间的信息相似度小于第三预设阈值的情况下，确定中间模型未达到收敛条件，从多个样本操作信息中获取下一样本，并将下一样本确定为当前样本。
在本实施例中，中间模型为神经网络模型未训练完成时的模型。目标模型的训练过程可以包括但不限于理解为，在玩家与玩家之间对战的整局游戏过程中，对玩家的环境参数，行动参数，结果参数进行采集。其中环境参数可以但不限于理解为游戏过程中玩家的操作被执行时所处的环境相关的参数，可以但不限于包括玩家在收到攻击时的位置信息、玩家在释放技能时的移动信息、玩家释放技能时周围的环境信息等等，这里不做多余限定。行为参数可以但不限于理解为游戏过程中玩家的操作对应的行为参数，可以但不限于包括释放一次高难度技能的具体操作指令、在接收到其他操作指令时应对的操作指令，等等。当前样本结果是与样本操作信息对应的操作匹配的、由第二操作对象执行的操作所对应的信息。本实施例中，当前样本结果也可以是指在玩家与玩家之间对战时的操作结果。
在本实施例中，在获取到环境参数、行为参数和结果参数的情况下，将当前样本输入到中间模型中，得到中间模型输出的中间操作信息，比对中间操作信息和当前样本结果之间的信息相似度，在信息相似度大于或等于第三预设阈值的情况下，确定中间模型达到收敛条件，并将中间模型确定为目标模型；在信息相似度小于第三预设阈值的情况下，确定中间模型没有达到收敛条件，从多个样本操作信息中获取其他样本，并对其他样本信息执行上述步骤以进行模型训练，直至中间模型达到收敛条件。通过结合环境参数、行为参数和样本结果进行模型训练，实现了提高模型训练的准确性的技术效果。
在本实施例中，可以但不限于采用损失函数来衡量目标模型的性能。将实际“激励值” 与模型预测行动产生的“激励值”的距离作为损失函数，例如如下公式(1)所示：
其中，L为LOSE的缩写，表示损失值。损失函数的计算过程实际是计算一个均方差，其中maxa′Q(s′,a′)函数表示目标值，例如在对抗类游戏中，虚拟对象可以扣血1滴或者2滴，但是若想要使得模型的性能更好，则会选取一个最大扣血值，这表示由用户控制的第一虚拟对象打出的最大伤害数值。函数的输出表示人工智能控制的第二虚拟对象打出的数值，将上述两个数值做差，然后求平方并除以2，即可得到损失值L。模型训练的目的是为了使损失值L等于0或者趋近于0，若L为0，则说明模型已达到完美状态，则模型训练完成，并将该模型确定为目标模型。在模型训练的过程中，将实际的激励值与模型预测行动产生的激励值的距离作为损失函数，通过反向传播对模型参数进行更新，以生成最终的目标模型。
本实施例中，使用非线性传播来逼近Q值的算法不稳定，很多情况下模型难以收敛，所以利用经验回放可以使得模型快速收敛。在传统客户端游戏中，在线对图像和操作指令等操作信息进行采集会产生较大的带宽负担，从而影响玩家的游戏体验。而云游戏本身就运行于服务端，直接利用服务端的资源对操作信息进行采集不会产生额外的带宽开销，采集成本较低。而且，云游戏服务集群本身拥有充足的GPU算力，模型训练过程中可以充分利用服务器的算力资源来降低运算成本。
在实施例中，在将多个操作信息输入目标模型之后，该方法还包括：
在操作信息为在一局云游戏的运行过程中采集到的图像信息的情况下，利用目标模型中的图像信息识别模块，对图像信息进行图像识别，得到处理后的操作信息；
将处理后的操作信息输入目标模型中的操作信息识别模块，得到第二目标操作信息。
通过本实施例，可以达到利用图像识别和模型训练得到第二目标操作信息的目的，从而实现了提高操作信息获取的准确性的技术效果。
在本实施例中，在生成目标模型之后，利用云游戏运行于服务端图像采集成本低的优势，采集玩家实时的游戏画面，通过模型运算得到人工智能的最接近玩家实际操作水平的操作指令或者操作模式。
需要说明的是，相比游戏运行于本地客户端，模型训练和图像采集都在云端完成，因此不会产生由于数据交互或者指令下达带来的额外延时，对于玩家而言，对手的反应会更及时，游戏体验更好，而且模型参数可以在线上进行实时更新。
在实施例中，获取在该局云游戏的运行过程中用户针对第一虚拟对象所产出的第一操作信息，并基于第一操作信息确定第二虚拟对象对应的第一操作模式，以及按照第一操作模式控制第二虚拟对象，包括：
获取在该局云游戏的运行过程中的第一时间段内用户针对第一虚拟对象所产出的第一操作信息，并基于第一操作信息确定第一操作模式，以及按照第一操作模式控制第二虚拟对象在第一时间段之后的行为操作。
在本实施例中，按照第一操作模式控制第二虚拟对象的行为操作可以但不限于理解为根据用户针对第一虚拟对象所产生的操作信息确定第一虚拟对象对应的操作水平为低难度水平(例如低于预设的难度阈值)，那么将第二虚拟对象对应的操作模式设置为低难度，从而控制第二虚拟对象在第一时间段后执行中低难度的操作指令。
通过本申请提供的实施例，能够达到基于时间段内的操作信息确定操作模式的目的，实现了提高操作模式的多样性的技术效果。
在实施例中，在按照第一操作模式控制第二虚拟对象在第一时间段之后的行为操作之后，该方法还包括：
获取在该局云游戏的运行过程中的、在第一时间段之后的第二时间段内用户针对第一虚拟对象所产出的第二操作信息，并基于第二操作信息确定第二操作模式，以及按照第二操作模式控制第二虚拟对象在第二时间段之后的行为操作。
通过本申请提供的实施例，可以达到灵活的控制第二虚拟对象的目的，实现了提高对虚拟对象控制的灵活性的技术效果。
在实施例中，按照第一操作模式控制第二虚拟对象，包括：
控制第二虚拟对象执行第一操作模式对应的至少一个第一操作指令；
按照第二操作模式控制第二虚拟对象，包括：控制第二虚拟对象执行第二操作模式对应的至少一个第二操作指令。
在本实施例中，可以但不限于根据不同操作模式控制虚拟对象执行对应的操作指令，例如根据操作信息判断玩家能够完成高难度的操作指令，则将第二虚拟对象执行的操作指令调整为较难释放的操作指令，从而实现了根据玩家针对第一虚拟对象的操作信息，灵活调整第二虚拟对象的具体操作指令的目的，达到了提高虚拟对象控制的灵活性的技术效果。
为便于理解，可以将上述虚拟对象的控制方法应用在具体的应用场景中，如图6所示，该方法可以包括如下步骤S602-S614：
步骤S602，检测玩家是否进入游戏连接。
步骤S604，根据玩家的等级、经验、胜率等等基础信息进行匹配对方玩家。
步骤S606，若匹配成功后，则进入到真人对战的游戏场景，即使得玩家控制的第一虚拟对象对战匹配的对方玩家控制的第三虚拟对象，完成游戏。
步骤S608，在玩家控制的第一虚拟对象与对方玩家控制的第三虚拟对象进行游戏时，利用实时图像信息和行动结果周期性地对玩家的行为操作等操作信息进行采集，例如大概每32ms采集一次，从实时图像中抓取数值，例如血量值、实时图像信息包括当前玩家的位置和玩家的行为操作，行动结果可以但不限于包括对战扣血值等等。
步骤S610，采集操作信息之后，将操作信息输入深度学习网络模型进行训练。
深度学习网络(Deep Q Network)模型的构建可以但不限于如图7中所示。状态参数702(state)表示玩家当前处于什么样的环境，比如两人对战的游戏场景，第一虚拟对象的目标位于什么位置等。行为参数704(action)表示第一虚拟对象针对目标采取的行为操作，例如往前移动或者踢一脚等。将这两种参数输入到Deep Q Network网络中，该网络输出一 个目标值(Q-value)。如果是两人对战的游戏场景中，使用扣血值作为目标的话，则使用Q-value作为最大扣血值，如果是其他游戏场景，则可以定义不同含义的目标。
神经网络是深度学习网络的基础，例如图8所示的神经网络模型包括一个输入层、三个卷积层、两个全连接层和一个输出层。卷积层的目的是为了提取图像中的信息，进行图像识别，比如识别虚拟对象位于图像的什么位置，或者从图像上抓取一些信息，比如说玩家控制的虚拟对象的最大胜率是多少等等。训练神经网络模型的目的在于，在确定第一虚拟对象处于某种状态的时候，由人工智能模拟的第二虚拟对象能够采取与该状态对应的策略。神经网络模型的训练需要设置一个优化目标，从而通过训练让神经网络模型输出符合需求的数值。神经网络模型的构建本质是构建一个公式来调整神经网络模型的参数，例如通过上面的公式(1)所示的损失函数来衡量神经网络模型的性能。
步骤S612-S614，若未匹配成功，则进入人工智能对战的游戏场景，采集由玩家控制的第一虚拟对象的实时画面，这里可以对每一帧图像都进行采集，并将采集的数据输入到已训练完成的神经网络模型进行运算，得到人工智能控制的第二虚拟对象的最佳反应，例如人工智能的操作指令或者操作模式等等。相对于游戏运行于本地客户端，模型训练和图像采集都在云端完成，因此不会产生由于数据交互或指令下达带来的额外延时，对于玩家而言，对手的反应会更及时，游戏体验更好，而且模型参数可以在线上进行实时更新，无需占用带宽下发到本地客户端。
在本实施例中，模型训练过程的流程可以如图9的训练时序图所示。用户客户端902将玩家输入的操作指令上传到服务器中的云游戏沙盒进程904，数据采集模块906从游戏沙盒进程904中抓取当前图像以及玩家操作指令、以及操作过后的扣血信息，完成图像、操作指令、激励值的采集，然后把图像和操作指令作为输入数据输入到Deep Q Network网络模型908中，对Deep Q Network网络模型908进行训练，利用反向传播对模型参数进行更新。
在本实施例中，玩家与人工智能的对战过程的流程可以如图10的时序图所示。用户客户端1002将玩家的操作指令输入云游戏沙盒进程1004，数据采集模块1006从云游戏沙盒进程1004中进行图像采集，提取玩家的当前游戏画面和操作指令，将提取的游戏画面和操作指令输入到训练好的Deep Q Network网络模型1008中。Deep Q Network网络模型1008返回人工智能的操作指令到云游戏沙盒进程1004中，云游戏沙盒进程1004接收到返回的操作指令，将操作指令输入到游戏进程中，使得人工智能根据该操作指令进行操作，然后将游戏对战画面返回到云游戏服务器中，云游戏服务器再将其返回给用户客户端1002进行显示。
通过上述实施例所述的方法，可以生成更符合人类思维的AI，优化AI的不合理部分，让玩家觉得AI像真人一样是鲜活有生命力的，从而更好地吸引玩家，增强玩家粘性。此外，通过上述实施例所述的方法还可以优化AI的匹配机制，更准确地评估玩家的游戏水平，从而匹配水平相当的对手及队友，保证游戏难度适中，让玩家获得更好的游戏体验，享受竞技的乐趣。
可以理解的是，在本申请的具体实施方式中，可能涉及到用户信息等相关的数据。当本申请以上实施例应用到具体产品或技术中时，需要获得用户许可或者同意，且相关数据的收集、使用和处理需要遵守相关法律法规和标准。
需要说明的是，对于前述的各方法实施例，为了便于描述，故将其都表述为一系列的步骤组合，但是本领域技术人员应该知悉，本申请并不受所描述的步骤顺序的限制，因为依据本申请，某些步骤可以采用其他顺序或者同时进行。其次，本领域技术人员也应该知悉，说明书中所描述的实施例均属于示例性实施例，所涉及的步骤和模块并不一定是本申请所必须的。
根据本申请实施例，还提供了一种用于实施上述虚拟对象的控制方法的虚拟对象的控制装置。如图11所示，该装置包括：
第一显示单元1102，用于在一局云游戏的运行过程中，显示参与该局云游戏的第一虚拟对象、以及第二虚拟对象，其中，第一虚拟对象为云游戏的用户所控制的虚拟对象，第二虚拟对象为人工智能控制的虚拟对象；
第一确定单元1104，用于获取在该局云游戏的运行过程中用户针对第一虚拟对象所产出的第一操作信息，并基于第一操作信息确定第二虚拟对象对应的第一操作模式，以及按照第一操作模式控制第二虚拟对象；
第一调整单元1106，用于在获取到用户针对第一虚拟对象所产出的第二操作信息的情况下，基于第二操作信息将第二虚拟对象对应的第一操作模式调整为第二操作模式，并按照第二操作模式控制第二虚拟对象，其中，第一操作信息不同于第二操作信息，第一操作模式不同于第二操作模式。
在实施例中，该装置还包括：
第二确定单元，用于在获取第一操作信息之后，确定第一操作模式的起始时间点；
第一获取单元，用于获取在起始时间点之后、在该局云游戏的运行过程中用户针对第一虚拟对象的多个操作信息；
第二获取单元，用于基于多个操作信息，确定第二操作信息。
在实施例中，第二获取单元包括：
第一确定模块，用于从多个操作信息中确定出与关键操作信息之间信息相似度大于或等于第一预设阈值的第一目标操作信息；
第二确定模块，用于在第一目标操作信息的数量大于或等于第二预设阈值的情况下，将第一目标操作信息确定为第二操作信息。
在实施例中，第二获取单元还包括：
第一输入模块，用于将多个操作信息输入目标模型，其中，目标模型为利用多个样本操作信息进行训练得到的、用于识别用户针对第一虚拟对象的操作信息的神经网络模型。
第二获取模块，用于获取目标模型输出的第二目标操作信息；
第三确定模块，用于在第二目标操作信息不同于第一操作信息的情况下，将第二目标操作信息确定为第二操作信息。
在实施例中，该装置还包括：
第四确定模块，包括第一获取子模块、第一输入子模块、第一确定子模块和第二确定子模块，用于在将多个操作信息输入目标模型之前，执行以下步骤，直至得到目标模型。
第一获取子模块，用于从多个样本操作信息中获取当前样本，其中，每个样本操作信息包含有当前环境参数、当前行为参数、以及当前样本结果，当前环境参数为样本操作信息对应的操作被执行时所处的环境相关的参数，当前行为参数为样本操作信息对应的操作对应的行为类型，当前样本结果为与样本操作信息对应的操作匹配的、由第二虚拟对象执行的操作所对应的信息。
第一输入子模块，用于将当前样本输入当中间模型，得到中间模型输出的中间操作信息，中间模型为所述神经网络模型未训练完成时的模型。
第一确定子模块，用于在中间操作信息与当前样本结果之间的信息相似度大于或等于第三预设阈值的情况下，确定中间模型达到收敛条件，并将中间模型确定为目标模型。
第二确定子模块，用于在中间操作信息与当前样本结果之间的信息相似度小于第三预设阈值的情况下，确定中间模型未达到收敛条件，从多个样本操作信息中获取下一样本，并将下一样本确定为当前样本。
在实施例中，该装置还包括：
第一识别模块，用于在将多个操作信息输入目标模型之后，在操作信息为在一局云游戏的运行过程中采集到的图像信息的情况下，利用目标模型中的图像信息识别模块，对图像信息进行图像识别，得到处理后的操作信息；
第二输入模块，用于将处理后的操作信息输入目标模型中的操作信息识别模块，得到第二目标操作信息。
在实施例中，第一确定单元1104包括：
第三获取模块，用于获取在该局云游戏的运行过程中的第一时间段内用户针对第一虚拟对象所产出的第一操作信息，并基于第一操作信息确定第一操作模式，以及按照第一操作模式控制第二虚拟对象在第一时间段之后的行为操作。
在实施例中，该装置还包括：
第四获取模块，用于在按照第一操作模式控制第二虚拟对象在第一时间段之后的行为操作之后，获取在该局云游戏的运行过程中的、在第一时间段之后的第二时间段内用户针对第一虚拟对象所产出的第二操作信息，并基于第二操作信息确定第二操作模式，以及按照第二操作模式控制第二虚拟对象在第二时间段之后的行为操作。
在实施例中，该第一确定单元1104包括第一控制模块，用于控制第二虚拟对象执行第一操作模式对应的至少一个第一操作指令；
该第一调整单元1106包括第二控制模块，用于控制第二虚拟对象执行第二操作模式对应的至少一个第二操作指令。
装置的具体实施方式可以参考上述虚拟对象的控制的实施例，在此不再赘述。
根据本申请实施例，还提供了一种用于实施上述虚拟对象的控制方法的电子设备。如 图12所示，该电子设备包括存储器1202和处理器1204，该存储器1202中存储有计算机程序，该处理器1204被设置为通过执行计算机程序以执行上述任一项方法实施例中的步骤。
在实施例中，上述电子设备可以位于计算机网络的多个网络设备中的至少一个网络设备。
本领域普通技术人员可以理解，图12所示的结构仅为示意性实施例，电子设备也可以是智能手机(如Android手机、iOS手机等)、平板电脑、掌上电脑以及移动互联网设备(Mobile Internet Devices，MID)、PAD等终端设备。图12并不对上述电子设备的结构造成限定。例如，电子设备还可包括比图12中所示更多或者更少的组件(如网络接口等)，或者具有与图12所示不同的配置。
在实施例中，存储器1202可用于存储软件程序以及模块，如本申请实施例中的虚拟对象的控制方法和装置所对应的程序指令和/或模块，处理器1204通过运行存储在存储器1202内的软件程序以及模块，从而执行各种功能应用以及数据处理，即实现上述的虚拟对象的控制方法。存储器1202可包括高速随机存储器，还可以包括非易失性存储器，如一个或者多个磁性存储装置、闪存、或者其他非易失性固态存储器。在一些实例中，存储器1202可进一步包括相对于处理器1204远程设置的存储器，这些远程存储器可以通过网络连接至终端设备。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。其中，存储器1202具体可以但不限于用于存储第一操作信息、第二操作信息等信息。作为一种示例，如图12所示，上述存储器1202中可以但不限于包括上述虚拟对象的控制装置中的第一显示单元1102、第一确定单元1104、第一调整单元1106。此外，还可以包括但不限于上述虚拟对象的控制装置中的其他模块单元，本示例中不再赘述。
在实施例中，上述的传输装置1206用于经由一个网络接收或者发送数据。上述的网络具体可包括有线网络及无线网络。在一个实例中，传输装置1206包括一个网络适配器(Network Interface Controller，NIC)，其可通过网线与其他网络设备与路由器相连，从而可与互联网或局域网进行通讯。在一个实例中，传输装置1206为射频(Radio Frequency，RF)模块，其用于通过无线方式与互联网进行通讯。
此外，上述电子设备还包括：显示器1208，用于显示上述第一操作信息、第二操作信息等；和连接总线1210，用于连接上述电子设备中的各个模块部件。
在其他实施例中，上述终端设备或者服务器可以是一个分布式***中的一个节点，其中，该分布式***可以为区块链***，该区块链***可以是由多个节点通过网络通信的形式连接形成的分布式***。其中，节点之间可以组成点对点(Peer To Peer，简称P2P)网络，任意形式的计算设备，比如服务器、终端设备等电子设备都可以通过加入该点对点网络而成为该区块链***中的一个节点。
需要说明的是，电子设备的计算机***仅是一个示例，不应对本申请实施例的功能和使用范围带来任何限制。
根据本申请的实施例，各个方法流程图中所描述的过程可以被实现为计算机软件程序。 例如，本申请的实施例包括一种计算机程序产品，其包括承载在计算机可读介质上的计算机程序和/或指令，该计算机程序和/或指令包含用于执行流程图所示的方法的程序代码。在这样的实施例中，该计算机程序可以通过通信部分从网络上被下载和安装，和/或从可拆卸介质被安装。在该计算机程序被中央处理器执行时，执行本申请的***中限定的各种功能。
根据本申请的一个方面，提供了一种计算机可读存储介质，用于存储计算机指令，计算机设备的处理器从计算机可读存储介质读取该计算机指令，处理器执行该计算机指令，使得该计算机设备执行上述各种实现方式中提供的方法。
在本实施例中，本领域普通技术人员可以理解，上述实施例的各种方法中的全部或部分步骤可以通过程序来指令终端设备相关的硬件来完成，该程序可以存储于一计算机可读存储介质中，存储介质可以包括：闪存盘、只读存储器(Read-Only Memory，ROM)、随机存取器(Random Access Memory，RAM)、磁盘或光盘等。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
上述实施例中的集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在上述计算机可读取的存储介质中。基于这样的理解，本申请的技术方案本质上或者说对相关技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在存储介质中，包括若干指令，用以使得一台或多台计算机设备(可为个人计算机、服务器或者网络设备等)执行本申请各个实施例方法的全部或部分步骤。
在本申请的上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
在本申请所提供的几个实施例中，应该理解到，所揭露的客户端可通过其它的方式实现。其中，以上所描述的装置实施例仅仅是示意性的，例如单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个***，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，单元或模块的间接耦合或通信连接可以是电性或其它的形式。
作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个设备，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
以上仅是本申请的部分实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本申请原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本申请的保护范围。

Claims (13)

  1. 一种虚拟对象的控制方法，包括：
    在一局云游戏的运行过程中，显示参与所述一局云游戏的第一虚拟对象、以及第二虚拟对象，其中，所述第一虚拟对象为所述云游戏的用户所控制的虚拟对象，所述第二虚拟对象为人工智能控制的虚拟对象；
    获取在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象所产出的第一操作信息，并基于所述第一操作信息确定所述第二虚拟对象对应的第一操作模式，以及按照所述第一操作模式控制所述第二虚拟对象；
    在获取到在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象所产出的第二操作信息的情况下，基于所述第二操作信息将所述第二虚拟对象对应的所述第一操作模式调整为第二操作模式，并按照所述第二操作模式控制所述第二虚拟对象，其中，所述第一操作信息不同于所述第二操作信息，所述第一操作模式不同于所述第二操作模式。
  2. 根据权利要求1所述的方法，其中，在所述获取所述第一操作信息之后，所述方法还包括：
    确定所述第一操作模式的起始时间点；
    获取在所述起始时间点之后、在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象的多个操作信息；
    基于所述多个操作信息，确定所述第二操作信息。
  3. 根据权利要求2所述的方法，其中，所述基于所述多个操作信息，确定所述第二操作信息，包括：
    从所述多个操作信息中确定出与关键操作信息之间信息相似度大于或等于第一预设阈值的第一目标操作信息；
    在所述第一目标操作信息的数量大于或等于第二预设阈值的情况下，将所述第一目标操作信息确定为所述第二操作信息。
  4. 根据权利要求2所述的方法，其中，所述基于所述多个操作信息，确定所述第二操作信息，包括：
    将所述多个操作信息输入目标模型，其中，所述目标模型为利用多个样本操作信息进行训练得到的、用于识别用户针对第一虚拟对象的操作信息的神经网络模型。
    获取所述目标模型输出的第二目标操作信息；
    在所述第二目标操作信息不同于所述第一操作信息的情况下，将所述第二目标操作信息确定为所述第二操作信息。
  5. 根据权利要求4所述的方法，其中，在所述将所述多个操作信息输入目标模型之前，所述方法包括：
    从所述多个样本操作信息中获取当前样本，其中，每个样本操作信息包含有当前环境参数、当前行为参数、以及当前样本结果，所述当前环境参数为所述样本操作信息对应的 操作被执行时所处的环境相关的参数，所述当前行为参数为所述样本操作信息对应的操作对应的行为类型，所述当前样本结果为与所述样本操作信息对应的操作匹配的、由所述第二虚拟对象执行的操作所对应的信息；
    将所述当前样本输入中间模型，得到所述中间模型输出的中间操作信息，所述中间模型为所述神经网络模型未训练完成时的模型；
    在所述中间操作信息与所述当前样本结果之间的信息相似度大于或等于第三预设阈值的情况下，确定所述中间模型达到收敛条件，并将所述中间模型确定为所述目标模型；
    在所述中间操作信息与所述当前样本结果之间的信息相似度小于所述第三预设阈值的情况下，确定所述中间模型未达到所述收敛条件，从所述多个样本操作信息中获取下一样本，并将所述下一样本确定为当前样本。
  6. 根据权利要求4所述的方法，其中，在所述将所述多个操作信息输入目标模型之后，所述方法还包括：
    在所述操作信息为在所述一局云游戏的运行过程中采集到的图像信息的情况下，利用所述目标模型中的图像信息识别模块，对所述图像信息进行图像识别，得到处理后的操作信息；
    将所述处理后的操作信息输入所述目标模型中的操作信息识别模块，得到所述第二目标操作信息。
  7. 根据权利要求1所述的方法，其中，所述获取在所述一局云游戏的运行过程中用户针对第一虚拟对象所产出的第一操作信息，并基于所述第一操作信息确定所述第二虚拟对象对应的第一操作模式，以及按照所述第一操作模式控制所述第二虚拟对象，包括：
    获取在所述一局云游戏的运行过程中的第一时间段内所述用户针对所述第一虚拟对象所产出的所述第一操作信息，并基于所述第一操作信息确定所述第一操作模式，以及按照所述第一操作模式控制所述第二虚拟对象在所述第一时间段之后的行为操作。
  8. 根据权利要求7所述的方法，其中，在按照所述第一操作模式控制所述第二虚拟对象在所述第一时间段之后的行为操作之后，所述方法还包括：
    获取在所述一局云游戏的运行过程中的、在所述第一时间段之后的第二时间段内所述用户针对所述第一虚拟对象所产出的所述第二操作信息，并基于所述第二操作信息确定所述第二操作模式，以及按照所述第二操作模式控制所述第二虚拟对象在所述第二时间段之后的行为操作。
  9. 根据权利要求1至8中任一项所述的方法，其中，
    所述按照所述第一操作模式控制所述第二虚拟对象，包括：控制所述第二虚拟对象执行所述第一操作模式对应的至少一个第一操作指令；
    所述按照所述第二操作模式控制所述第二虚拟对象，包括：控制所述第二虚拟对象执行所述第二操作模式对应的至少一个第二操作指令。
  10. 一种虚拟对象的控制装置，包括：
    第一显示单元，用于在一局云游戏的运行过程中，显示参与所述一局云游戏的第一虚 拟对象、以及第二虚拟对象，其中，所述第一虚拟对象为所述云游戏的用户所控制的虚拟对象，所述第二虚拟对象为人工智能控制的虚拟对象；
    第一确定单元，用于获取在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象所产出的第一操作信息，并基于所述第一操作信息确定所述第二虚拟对象对应的第一操作模式，以及按照所述第一操作模式控制所述第二虚拟对象；
    第一调整单元，用于在获取到在所述一局云游戏的运行过程中所述用户针对所述第一虚拟对象所产出的第二操作信息的情况下，基于所述第二操作信息将所述第二虚拟对象对应的所述第一操作模式调整为第二操作模式，并按照所述第二操作模式控制所述第二虚拟对象，其中，所述第一操作信息不同于所述第二操作信息，所述第一操作模式不同于所述第二操作模式。
  11. 一种计算机可读的存储介质，包括存储的程序，其中，所述程序可被终端设备或计算机运行时执行所述权利要求1至9任一项中所述的方法。
  12. 一种计算机程序产品，包括计算机程序和/或指令，其中，该计算机程序和/或指令被处理器执行时实现权利要求1至9任一项中所述方法的步骤。
  13. 一种电子设备，包括存储器和处理器，其中，所述存储器中存储有计算机程序，所述处理器被设置为通过执行所述计算机程序执行所述权利要求1至9任一项中所述的方法。
PCT/CN2023/129848 2022-11-22 2023-11-06 虚拟对象的控制方法、装置和存储介质及电子设备 WO2024109528A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211466604.6 2022-11-22
CN202211466604.6A CN116943204A (zh) 2022-11-22 2022-11-22 虚拟对象的控制方法、装置和存储介质及电子设备

Publications (1)

Publication Number Publication Date
WO2024109528A1 true WO2024109528A1 (zh) 2024-05-30

Family

ID=88457020

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/129848 WO2024109528A1 (zh) 2022-11-22 2023-11-06 虚拟对象的控制方法、装置和存储介质及电子设备

Country Status (2)

Country Link
CN (1) CN116943204A (zh)
WO (1) WO2024109528A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116943204A (zh) * 2022-11-22 2023-10-27 腾讯科技（深圳）有限公司 虚拟对象的控制方法、装置和存储介质及电子设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100279762A1 (en) * 2009-04-30 2010-11-04 Electronics And Telecommunications Research Institute Apparatus and method for adjusting difficulty level of game
WO2013180322A1 (ko) * 2012-05-30 2013-12-05 주식회사 쏘그웨어 게임 사용자 수준에 따른 ｎｐｃ ａｉ제공방법
CN109091869A (zh) * 2018-08-10 2018-12-28 腾讯科技（深圳）有限公司 虚拟对象的动作控制方法、装置、计算机设备及存储介质
CN111450533A (zh) * 2020-03-31 2020-07-28 腾讯科技（深圳）有限公司 虚拟场景中的虚拟对象控制方法、装置、终端及存储介质
CN116943204A (zh) * 2022-11-22 2023-10-27 腾讯科技（深圳）有限公司 虚拟对象的控制方法、装置和存储介质及电子设备

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100279762A1 (en) * 2009-04-30 2010-11-04 Electronics And Telecommunications Research Institute Apparatus and method for adjusting difficulty level of game
WO2013180322A1 (ko) * 2012-05-30 2013-12-05 주식회사 쏘그웨어 게임 사용자 수준에 따른 ｎｐｃ ａｉ제공방법
CN109091869A (zh) * 2018-08-10 2018-12-28 腾讯科技（深圳）有限公司 虚拟对象的动作控制方法、装置、计算机设备及存储介质
CN111450533A (zh) * 2020-03-31 2020-07-28 腾讯科技（深圳）有限公司 虚拟场景中的虚拟对象控制方法、装置、终端及存储介质
CN116943204A (zh) * 2022-11-22 2023-10-27 腾讯科技（深圳）有限公司 虚拟对象的控制方法、装置和存储介质及电子设备

Also Published As

Publication number Publication date
CN116943204A (zh) 2023-10-27

Similar Documents

Publication Publication Date Title
CN108888958B (zh) 虚拟场景中的虚拟对象控制方法、装置、设备及存储介质
CN111744201B (zh) 视频游戏中的自动玩家控制接管
US20220410007A1 (en) Virtual character interaction method and apparatus, computer device, and storage medium
WO2024109528A1 (zh) 虚拟对象的控制方法、装置和存储介质及电子设备
KR20210005603A (ko) 인공 지능을 사용하는 음성 지원 시스템
US11712630B2 (en) Dynamic interfaces for launching direct gameplay
US20150367238A1 (en) Game system, game apparatus, a method of controlling the same, a program, and a storage medium
CA2910655A1 (en) Information processing apparatus, control method and program
CN113069767B (zh) 虚拟互动方法、装置、终端和存储介质
JP7352368B2 (ja) ゲーム操作学習プログラム、ゲームプログラム、ゲームプレイプログラム、及びゲーム操作学習方法
CA2872130A1 (en) Information processing apparatus, rendering apparatus, method and program
JP5841110B2 (ja) 仮想環境における通信セッションの提示
JP6639540B2 (ja) ゲームシステム
EP3766553A1 (en) User interaction selection method and apparatus
JP4778742B2 (ja) ゲームシステム、ゲームプログラム、及び記録媒体
CN112288093A (zh) 使用机器学习来增加或降低击败视频游戏对手的难度级别
JP2007020843A5 (zh)
JP2022130494A (ja) コンピュータシステム
CN108905201A (zh) 一种模拟操作的游戏算法***及方法
Hirose et al. Robotic gaming companion to facilitate social interaction among children
JP2020062429A (ja) ゲームシステム
KR20210032838A (ko) 감정을 이용한 게임 장치 및 방법
WO2024051368A1 (zh) 信息显示方法、装置和存储介质及电子设备
KR102343359B1 (ko) 친구 감정 표정을 이용한 게임의 에너지 충전 장치 및 방법
JP6997386B2 (ja) ゲーム関連サービス提供プログラムならびにサーバ装置